Thomas 1998

INTERNATIONAL JOURNAL FOR NUMERICAL AND ANALYTICAL METHODS IN GEOMECHANICS, VOL.
22, 75—95 (1998)
THREE-DIMENSIONAL HEAT, MOISTURE AND AIR

TRANSFER IN UNSATURATED SOILS
H. R. THOMAS*, S. W. REES AND N. J. SLOPER

Queen’s Buildings, PO Box 917, Cardiff, CF2 1XH, Wales
SUMMARY
A new three-dimensional numerical model of coupled heat, moisture and air transfer in unsaturated soil is
presented. In particular, the model accommodates moisture transfer in the form of liquid and vapour flow
and heat transfer arising from conduction, convection and latent heat of vaporization. The bulk flow of dry
air and the movement of air in a dissolved state are also included. The theoretical basis of the model, the
finite element solution of the spatial terms and finite difference solution of the temporal terms are briefly
presented. Attention is focused on the verification of the new numerical solution. This is achieved via
comparisons with independent solutions of heat, moisture and air transfer in an unsaturated soil. The
physical problem considered includes the highly non-linear hydraulic properties of sand. Thermal conduct-
ivity is also included as a function of soil moisture content. Excellent correlation of results is shown thus
providing confidence in the new model.
The new model is also applied to a number of test cases which illustrate the need for the development of
a model which can fully include three-dimensional behaviour. In particular, three applications are presented
each increasing in complexity. The first application illustrates three-dimensional heat transfer. This particu-
lar application is verified against existing commercial finite element software. Subsequent applications serve
to illustrate how the coupled processes of heat moisture and air transfer combine to yield three-dimensional
problems even within a simple geometric domain. Visualization of three-dimensional results is also
addressed. ( 1998 by John Wiley & Sons, Ltd.
Int. J. Numer. Anal. Meth. Geomech., Vol. 22, 75—95 (1998)
(No. of Figures: 7 No. of Tables: 2 No. of Refs: 21)
Key words: three-dimensional; heat; moisture; air; transfer; unsaturated; soil
1. INTRODUCTION
The demand for an improved understanding of the thermo-hydraulic behaviour of unsaturated
soils arises from a number of sources, in particular, from recent developments within the fields of
civil and environmental engineering. For example, some of the more recent applications include
studies of the energy efficiency of buildings through analysis of the thermal performance of
ground floor slabs, and analysis of moisture migration under thermal gradients set-up in new
techniques for remediation of contaminated land. A further area of application concerns the
disposal of high-level nuclear waste. This is a subject of major international importance and is
viewed as a particularly significant research application of the current work.
*Correspondence to H. R. Thomas, Queen’s Buildings, PO Box 917, Cardiff, CF2 1XH, Wales
Tel:#44 (01222) 874279 Fax:#44 (01222) 874597 E-mail: thomashr@cf.ac.uk
CCC 0363—9061/98/020075—21$17.50 Received 25 July 1996

( 1998 by John Wiley & Sons Ltd. Revised 14 May 1997
76 H. R. THOMAS E¹ A¸.
Considering further the disposal of nuclear waste. One method under consideration is that of
deep geological burial, where the waste will be placed in a repository and a multibarrier
containment concept will be employed to mitigate the transmission of radionuclides to the
biosphere. Within the context of this concept the accurate prediction of the transfer of heat
generated by the high level waste, the corresponding migration of moisture within the engineered
barrier system and host material, and consequent changes in air pressure are viewed as important.
Assessment of the overall performance of the system will depend on a clear understanding of these
processes amongst others.1
The design and performance assessment of barriers proposed to contain nuclear waste in
underground repositories will ultimately require full-scale analyses which take into account
discontinuities in the host material, which may result in zones of distinctly varying hydraulic
properties, and more local heterogeniety caused by the drilling process. Detailed consideration of
the significance of the geometrical configuration (possibly non-symmetric) will also be required.
To this end, this paper presents the development of a three-dimensional model that is capable of
describing fully coupled transfer of heat, moisture and air within the context of this type of
practical engineering problem. The model is only applicable to non-deformable soil.
The theoretical formulation presented is based upon an approach developed by Thomas and
Samsom.1 Validation of the formulation has been considered previously in some detail. For
example, a range of comparisons between numerical and experimental results has been provided
by King.2 The potential-based approach employed3 differs from the moisture content/temper-
ature4 formulation more commonly adopted and the advantages of this are described elsewhere.5
Particular consideration has recently been given to validation of the air phase which is included in
the current model.2,6,7 In view of the extensive work carried out previously, the basic theoretical
formulation used here is taken to be of established validity.
The current work focuses on the development of a new three-dimensional model based on
established theory and the necessary verification of the new computer software. The development
of a numerical solution of the problem is briefly described. The finite element approach adopted
was originally suggested by Zeinkiewicz and Taylor.8 Verification of the new model is achieved
by a comparison of results with independent solutions.1 This is followed by a number of
applications of heat, moisture and air transfer through a small cubic domain and non-deformable
sand. In particular, the applications serve to illustrate how three-dimensional behaviour can
easily arise within the context of the coupled processes under consideration. The practical
application of the new model gives rise to difficulties related to the need for visualisation of the
three-dimensional numerical results. This problem is also addressed.
2. THEORY
The theoretical formulation employed considers flow through unsaturated soil as a three-phase
system comprising solid particles, liquid and gas. The gas phase consists of a mixture of dry air
and water vapour and the liquid phase is considered to be water which may contain dissolved air.
The governing equations for moisture flow, as a combination of liquid transfer and vapour
transfer, dry air flow and heat flow are briefly presented below. The governing equations are
solved for the dependent variables capillary potential, temperature and air pressure. A more
complete description of the theoretical formulation has been presented previously1 and, therefore,
only a brief summary of the formulation is included here.
Int. J. Numer. Anal. Meth. Geomech., Vol. 22, 75—95 (1998) ( 1998 by John Wiley & Sons, Ltd.
THREE-DIMENSIONAL HEAT, MOISTURE AND AIR TRANSFER 77
2.1. The governing differential equation for moisture flow

A generalized form of Darcy’s law is employed for non-isothermal liquid flow in an unsaturated
soil. The unsaturated hydraulic conductivity is introduced as a function of moisture content and
temperature by a relationship with the dynamic viscosity of water.9 The surface energy approach
is used to express the capillary potential as a function of moisture content and temperature.10
Through the conservation of mass for liquid flow and employing the principle of partial
derivatives the following equation for liquid flow is obtained:
L( L¹
C #C "+[K ( +(]#+[K +P ]#+[K ( +z]!o E (1)
L ( Lt LT Lt L L! ! L 1
where ( is the capillary potential, P is the air pressure and ¹ is the temperature. + is the vector
!
differential operator which maintains full three dimensionality throughout the formulation. All
terms are fully defined in the notation appendix.
The two effects of vapour pressure gradient and bulk flow of air are assumed to create vapour
flow. These effects are related by using conservation of mass for vapour flow and assuming that
liquid water is in equilibrium with water vapour.3 The effects of dry air velocity are introduced
from Darcy’s law for multiphase flow ignoring gravitational effects.11 By taking into account the
above effects and through appropriate substitution the following equation for vapour transfer is
produced:
L( L¹
C #C "+[K +(]#+[K +¹ ]#+[K +P ]
V( Lt VT Lt V( VT V! !
#» +(#» +¹#o E (2)

V( VT 1
The governing equations for liquid (1) and vapour flow (2) are combined to yield the following
equation for moisture flow
L( L¹
C #C "+[K +(]#+[K +¹ ]#+[K +P ]
(( Lt (T Lt (( (T (! !
#» +(#» ¹#+[K +z] (3)

(( (T L(
2.2. The governing differential equation for dry air flow

The isothermal transfer of dry air is assumed to be caused by two effects; the bulk flow of dry air
and the transfer of dissolved air in liquid water.12 The dissolved air is evaluated from Henry’s
coefficient of solubility13 and the flow of dry air by an application of the principle of conservation
of mass. The dry air and water vapour are considered as ideal gases14 and Dalton’s law of partial
pressure is applied. The resulting equation is differentiated with respect to time and yields the
governing differential equation for dry air flow as
L( L¹ LP
C #C #C !"+[K +(]#+[K +P ]
!( Lt !T Lt !! Lt !( !! !
C D
LK
#» +(#» +¹#» +P #o H + (4)
!( !T !! ! $! # Lz
( 1998 by John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech., Vol. 22, 75—95 (1998)
2.3. The governing differential equation for heat flow

The effects of conduction, convection and latent heat of vaporization are included in the
equation for heat flow. Convection results from the movement of liquid water, water vapour
(from the vapour pressure gradient) and dry air. Latent heat is associated with the movement of
water vapour from the vapour pressure gradient and as part of bulk flow of air. The principle of
energy conservation is applied and the result is simplified to give the following governing
differential equation for heat transfer:
L( L¹ LP
C #C #C !"+[K +(]#+[K +¹ ]#+[K +P ]
T( Lt TT Lt T! Lt T( TT T! !
#» ( +(#» +¹#» +P #C (¹!¹ )+[K +z] (5)
T TT T! ! 1- 3 (
Together equations (3)—(5) define the complete theoretical formulation describing the coupled
processes of moisture, air and heat flow in a three-dimensional unsaturated soil. These equations
contain only three unknowns; the dependent variables ( (capillary potential), ¹ (temperature)
and P (pore air pressure).
!
3. THE NUMERICAL SOLUTION OF THE GOVERNING PARTIAL DIFFERENTIAL

EQUATIONS
Due to the complexity of the theoretical formulation, numerical approximations are required to
achieve the simultaneous solutions to the governing differential equations and the salient
concepts of the numerical formulation employed have been described in detail elsewhere.8,15
A fully coupled spatial solution is achieved through the simultaneous solution of the three
equations employing the finite element method, in particular, the Galerkin weighed-residual
approach. The domain is discretised using three dimensional 20-node isoparametric ‘brick’
elements. Serendipity shape functions are used to approximate the unknown primary variables:
CD CD CD
( ( K (
20 s
¹ + ¹K " + N ¹ (6)
s s
P PK s/1 P
! ! !s
Figure 1. Boundary flux on a 20 noded isoparametric ‘brick’ element
where (ª , ¹K and PK are the approximate values of capillary potential, temperature and pore air
!
pressure, respectively, and N are the standard serendipity shape functions.8
s
After application of the Gauss-Green divergence theorem, which introduces surface integrals
that accommodate the summation of fluxes normal to the boundary surface and by substitution
of the derivatives of the shape function approximations given in equations (6) the three governing
equations (3)—(5) may be recast as shown below:
The spatial discretization of the governing equation for moisture transfer yields:
P)% [K +N3 +N4 ] d)%W4#P) [K T +N3 +N4 ] d)%T4#P) [K !+N3+N4 ] d)% P!4
((
%
(
%
(
LW
P) 3 4 4 P) 3 T 4 4 P) 3 4 Lt
! [N » +N ] d)%W ! [N » +N ] d)%T # [N C N ] d)% 4
%
((
%
(
%
((
LT
P) 3 T 4 Lt P) -! 3
%
(
P 3 1 -/ 7/
# [N C N ] d)% 4# [K +N +z] d)%! N o [v #v ] d!%"0
% !%
(7)
The spatial discretization of the governing equation for heat transfer yields:
P) [KT +N3+N4 ] d)%W4#P) [KTT+N3+N4 ] d)%T4#P) [KT!+N3+N4 ] d)%P!4

%
(
% %
P) 3 T 4 4 P) 3 TT 4 4 P) 3 T! 4 !4
! [N » +N ] d)%W ! [N » +N ] d)%T ! [N » +N ] d)%P
%
(
% %
LW LT LP
P) 3 T 4 Lt P) 3 TT 4 Lt P) 3 T! 4 Lt
# [N C N ] d)% 4# [N C N ] d)% 4# [N C N ] d)% !4
%
(
% %
P) 1- % - 3 P 3 -)1 )2 )3 )4 )5
# [C (¹!¹)K +N +z] d)%! N [F #F #F #F #F ] d!%"0
(
!%
(8)
The spatial discretization of the governing equation for dry air transfer yields:
P) [K! +N3+N4 ] d)%W4#P) [K!4+N3+N4 ] d)%P!4!P) [N3»! +N4 ] d)% W4

%
(
% %
(
LW
P) 3 !T 4 4 P) 3 !! 4 !4 P) 3 ! 4 Lt
! [N » +N ] d)% T ! [N » +N ] d)%P # [N C N ] d)% 4
% % %
(
LT LP
P) 3 !T 4 Lt P) 3 !! 4 Lt P) $! # - 3
# [N C N ] d)% 4# [N C N ] d)% !4# [o H K +N +z] d)%
% % %
(
P 3 $! &/ $/
! N o [v #v ] d!%"0
!%
(9)
Equations (7)—(9) may be combined and written more conveniently in matrix notation as
C DG H C DG H G H
K(( K( K( W C(( C( — W0 J(
T ! 4 T 4
K ( K K T # C ( C C T0 # J "M0N (10)
T TT T! 4 T TT T! 4 T
K( K K P C( C C P0 J
! !T !! !4 ! !T !! !4 4
The vector J is composed of gravity and flux boundary terms. Equation (10) may be rewritten as
L/
K/#C #J"0 (11)
Lt
The boundary fluxes are evaluated as a normal flow acting on an element face using a method
similar to that suggested by Irons and Ahmad.16 This method has been adapted from the
structural terms of force and load16 to represent the appropriate flux terms encountered within
the heat, moisture and air transfer model.
The elemental area dA is evaluated by finding the magnitude of the cross product of the two
in-plane direction vectors:
dA"Dn]gD dm dg (12)
Where g and n are local co-ordinates. The nodal components are then given by
`1 `1
k"
i P~1 P~1 Ni jkª i Dm]gD dm dg (13)
where the subscript i refers to the nodel value, ˆ refers to the approximate value and j is the flux
component which is assumed to be always acting normal to the surface. The unit vector kª ,
therefore, is given by
n]g
kª " (14)
Dn]gD
and therefore the nodal components are given by
`1 `1
k"
i P~1 P~1 Ni jn]gdm dg (15)
To obtain the temporal solution to equation (11) a fully implicit mid-interval backward difference
algorithm is employed. For a mid-interval backward difference scheme equation (11) may
rearranged and simplified to give
C D C D
Cn`1@2 ~1 Cn`1@2/n
/n`1" Kn`1@2# · !Jn`1@2 (16)
*t *t
A solution for /n`1 is obtained by calculating the coefficients K, C and J at the mid-time interval
n#1/2. This is achieved using a predictor—corrector iterative solution procedure. A converged
solution is assumed to have been achieved when the difference between successive iterations falls
below a specified tolerance.
4. MODEL VERIFICATION
The validity of the theoretical formulation previously adopted for a two-dimensional model for
heat, moisture and air transfer1 has been shown through comparison of numerical results with
experimental results.2,6,7 Therefore, verification of the new three-dimensional formulation was
considered through a comparison of numerical results with a two-dimensional flow analysis of
fully coupled heat, moisture and air transfer in unsaturated soil.1 Full validation of the three-
dimensional model against experimental data is not possible at the current time due to a lack of
appropriate experimental data.
The analysis used for verification of the three-dimensional model illustrates the effects of
thermal drying, as a result of elevated temperatures, hydration, due to recovery of ground water
level, and the influence of elevated air pressures. The original two-dimensional analysis con-
sidered a 100 mm column of sand comprised of fifty uniform isoparametric eight noded elements
with dimensions 2 mm]20 mm as shown in Figure 2(a). The new three-dimensional model was
used to re-analyse this problem. A three-dimensional mesh was developed adopting similar
discretization using fifty, 20-node isoparametric elements, of dimensions 2 mm]20 mm]20 mm,
as shown in Figure 2(b). The material employed is unsaturated natural Leighton Buzzard medium
sand with material parameters independently measured.17,18 The material properties required in
the numerical model are defined in Table I.
A uniform set of initial conditions are employed throughout the domain; a capillary potential
of !0.5 m, representing an initial moisture content of 1·7 per cent, air pressure is set at an
atmosphere value of 100 kPa and the initial temperature is 293 K. Dirichlet boundary conditions
were applied to the top and bottom faces of the column. The lower face of the column was subject
to an elevated temperature of 50°C to represent heat generated by a waste source. The upper
surface was prescribed at a capillary potential of !0·3 m simulating infiltration. Air pressure at
the upper surface is at an initial value of 100 kPa and is allowed to increase to a prescribed
threshold value of 102 kPa. This enables the air pressure to increase in the sample to this
threshold value at such time it is allowed to escape. At this time the capillary potential is
decreased to the initial value of !0·5 m. An initial timestep size of 10 s is employed which is
allowed to increase to a maximum size of 900 s. This increase in timestep size is controlled by
a factor which is a function of the number of iterations required to provide a converged solution.
Results are shown in Figures 3(a) and 3(b). Figure 3(a) shows temperature distribution at
various times taken at sampling points spaced at 4 mm intervals. The sampling points have been
taken from the midsides of each element face and have proved to be identical at corresponding
depths to those achieved by Thomas and Sansom.1 As would be expected the temperature
Figure 2. Verification—Column finite element meshes: (a) 2D; (b) 3D
Figure 3. Verification—Transient variations with depth: (a) Temperature; (b) Capillary potential
throughout the domain steadily increases due to the effect of the elevated temperature boundary
until a steady-state condition is achieved at a time in excess of 5 h. Figure 3(b) shows capillary
potential distributions shown at various times taken at the same sampling points as mentioned in
the previous Figure 3(a). Again results from Thomas and Sansom1 were found to be identical at
corresponding depths. A full description of the behaviour exhibited by the capillary potential
profiles is provided by Thomas and Sansom1 and for conciseness is not repeated here.
It is encouraging to note the high degree of agreement between the new three-dimensional
model and an independent two-dimensional model. This analysis is one of many similar exercises
Figure 3 (Continued)
that have been carried out in order to verify the new model. In total a series of some 18
verification exercises have been completed yielding a high degree of confidence in the three-
dimensional model.
5. THREE-DIMENSIONAL APPLICATIONS
Following verification of the new model a number of applications have been considered to
investigate the three-dimensional behaviour of heat, moisture and air flow through unsaturated
soil. Some of the applications undertaken in this respect are presented below. All applications use
the same material, Leighton Buzzard medium sand (described above), and the same spatial
domain.
5.1. Finite element discretisation

The domain is shown in Figure 4(a) and consists of 64 regular cubic elements of dimension
12·5]12·5]12·5 mm resulting in an overall cubic domain of dimension 50]50]50 mm. The
domain has been restricted to this size in the first instance due to the high computational (CPU)
time taken to obtain a solution.
Using a three-dimensional domain can lead to difficulty for the reader regarding the interpreta-
tion of results and as a result this paper will endeavour to provide a consistent nomenclature.
Figures 4(b) and 4(c) show definitions for plane cross-sections, cube faces, one-dimensional cross
sections and the co-ordinate system. The vertical dimension is labelled Z for a consistent
nomenclature with elevation (z) and therefore co-ordinates are given in the order X, ½, Z through-
out. A plane across the ½ and Z dimensions at a distance of X"0 mm will be known as X0.
Figure 4. Cube domain: (a) Finite element mesh; (b) Faces; (c) 1D cross sections
Similarly a plane in the X and ½ dimensions at a vertical height of Z"50 mm (i.e., the top surface
of the domain) will be known as Z50. One-dimensional cross-sections are taken at the centre of
a cube face through the domain as shown in Figure 4(c). For instance X—X provides a spatial
variation in the X dimension through the centre of the ½Z plane (i.e., ½"Z"25 mm).
An additional factor that can require clarification arrives from the use of capillary potential (()
as the primary variable describing unsaturated moisture transfer. The units of this variable are
that of negative head (!( m), therefore, the larger the absolute value of capillary potential the
lower the soil moisture content.
5.2. Application 1—Thermal application

The first application only considers heat conduction and the problem is run uncoupled with
values of capillary potential (() and air pressure (P ) held constant at values of —0.5 m and
!
100 kPa (+1 atm), respectively. Initial temperatures throughout the domain are set uniformly at
a value of 20°C. Dirichlet boundary conditions are set at all the faces of the domain and provide
thermal gradients in the three dimensions. Faces X0, ½0 and Z0 are set at an elevated
temperature of 60°C with faces X50, ½50 and Z50 held at the initial temperature of 20°C. For
temporal changes a variable time-stepping procedure is adopted. An initial timestep size of 10 s is
employed which is allowed to increase, subject to certain convergence criteria, to a maximum
value of 50 s.
Results are presented in Figures 5(a) to 5(c). Figures 5(a) and 5(b) show contours of temperature
taken at times of 10 s and 600 s, respectively, through plane section ½12·5 (i.e. a ZX plane at
a distance of ½"12·5 mm). Figure 5(a) shows temperatures are highest adjacent to the prescribed
boundary on faces X0 and Z0. Figure 5(b) shows the same plane (½12·5) taken at the equilibrium
position of 600 s. It may be seen that the overall domain has, as would be expected, increased in
temperature. It can also be seen that the mid-temperature, the average of the two boundary
conditions at 40°C, is offset from the centre of the X0-X50 and Z0-Z50 gradients. This can be
attributed to the ½0-½50 gradient in the ½ direction increasing the overall temperature at the
cross-section position. Figure 5(c) shows plane sections ½25 and Z25 at the equilibrium time of
600 s. The Figure shows temperature distribution through the domain with red representing
a temperature of 60°C and blue a temperature of 20°C. This figure qualitatively shows the thermal
distribution through the cube is clearly three dimensional.
A further verification exercise was conducted using a commercial finite element software
package for three dimensional heat flow.19 The identical spatial domain used in Application 1
was modelled using PAFEC. The PAFEC software only considered heat conduction and
therefore the material properties required for this software were those of specific heat capacity
and thermal conductivity. The value adopted for specific heat capacity was 800 J/kgK (Table I)
and the value for thermal conductivity of 1·7126 W/mK. The constant value for thermal
conductivity was achieved by substituting the constant initial capillary potential value of !0·5 m
into the t v h relation (Table I) yielding a volumetric liquid content of 1·7 per cent and
substituting this value into the j v h relation (also given in Table I).
# 1
Results for this analysis are shown in Figure 5(d). The figure makes a comparison of temper-
ature profiles, across section X-X (see Figure 4(c)), obtained by the new three-dimensional model
and by the PAFEC model at equilibrium. This figure clearly demonstrates the high level of
quantitative agreement between the thermal response of the new and alternative models and
confidence in the validity of the new model has been further enhanced.
Figure 5(a). Application 1—Thermal distribution for plane ½12·5 at 10 s. (b) Application 1—Thermal distribution for
plane ½12·5 at 600 s. (c) Application 1—3D temperature distribution for whole domain. (d) Application 1—A comparison
of the equilibrium profile at section X-X between the new model and PAFEC
5.3. Application 2—Heat and moisture transfer application

The second application seeks to show effects, through three separate simulations, of gravity,
a moisture gradient and a thermal gradient operating in the Z, X and ½ dimensions, respectively.
The three simulations are run as fully coupled heat and moisture transfer analysis with air
pressure assumed to be a constant 100 kPa. Simulation 1 considers the effect alone of a moisture
gradient on the domain. Simulation 2 retains the moisture gradient set up in Simulation 1 and
Table I. Soil properties for Leighton Buzzard Sand
Property/relationship Value/equation
Density (o ) 2700 kg/m3

4
Specific heat capacity (C ) 800 J/kgK
14
Intrinsic permeability (k) 9·6]10~11 m2
Porosity (n) 0·389
Soil water retention curve (( v h) ( "!2·41!0·002h~1.75 m 0)h )0·2
3 - -
( "[SM (n!h )]Q1 m 0·2)h )n
3 - -
where SM "1·101416]10~26 and QM "0·0226249
Hydraulic conductivity v
C A BD
volumetric liquid content (K v h ) h h 2
- - K "1·5]10~-0 exp 28·061 -!12·235 - m/s
- n n
Thermal conductivity v volumetric liquid j "0·256#2·458[1!exp(!22·94h )] W/mK
# -
content (j v h )
# -
Relative air permeability v volumetric liquid k "(1!S )2 (1!S(2`u)/u) m/s
3! % %
content (k v h ) (h /n)!S
3! - where S " - 3
% 1!S
3
Table II. Summary of boundary conditions used by the simulations in the heat and mass
transfer analysis
Dirichlet boundary conditions
Simulation Gravity Capillary potential (() Temperature (¹ )
1 ] !0·5 m over X0 —
!0·3 m over X50
2 @ !0·5 m over X0 —
!0·3 m over X50
3 @ !0·5 m over X0 20°C over ½0
!0·3 m over X50 60°C over ½50
also includes the effects of gravity on the domain. Simulation 3 retains the moisture gradient and
gravitational potential and additionally includes a thermal gradient. All simulations are started
with uniform initial conditions of !0·5 m and 20°C for capillary potential and temperature,
respectively. All simulations are subjected to Dirichlet capillary potential boundary values of
!0·3 m on face X50 and !0·5 m on face X0, inducing a moisture gradient in the X direction.
Simulation 2 includes gravitational effects inducing a moisture gradient in the Z (vertical)
direction. Simulation 3 is subject to an elevated temperature boundary values of 60°C on face ½50
and 20°C on face ½0, inducing a thermal gradient in the Y direction. Boundary conditions are
summarised in Table II. Unless stated in Table II all other boundaries are non-conducting.
The contribution of each gradient is highlighted as the simulations progress through this
analysis. Figures 6(a) to 6(c) present comparisons of the results achieved by each simulation.
Figure 6(a) shows the transient variation of capillary potential (clear markers) and temperature
(solid markers) at point 25, 12·5, 25 (X, Z, ½, i.e. a point offset vertically from the centre of the
domain) for each simulaton. The effect of gravity on the moisture gradient in Simulation 1 can be
Figure 6(a). Application 2—Transient variation of temperature and capillary potential at point (25,12·5,25). (b) Applica-
tion 2—Temperature and capillary potential profiles along section X-X. (c) Application 2—Capillary potential contours
after 5000 s on plane Z25
seen by comparison with Simulation 2, which shows the capillary potential variation with time
including gravitational potential in the analysis. Initially, gravity seems to cause little effect on the
capillary potential distribution but as time increases the influence of gravity on the capillary
potential at the sample point becomes more evident as a new gradient is established in the vertical
Z dimension.
The influence of the addition of a thermal gradient in Simulation 3 is shown by the curve
‘Simulation 3’ in Figure 6(a). A curve is also provided (represented by solid markers) of the
corresponding temperature increase at the point. The elevated temperature has the clear effect of
increasing the soil moisture content (i.e. the capillary potential becomes less negative) at this
sample point in comparison with the previous moisture content values. The increase is attributed
to the hotter temperature at face ½50 driving moisture away thereby increasing the moisture
content elsewhere in the cooler regions of the domain. This effect is further highlighted by Figures
6(b) and 6(c).
Figure 6(b) shows section X!X (i.e. ½ and Z constant at 25 mm) at two times of 500 and
100,000 s (+27·7 h) for Simulation 2 (moisture and gravity gradients) and Simulation 3 moisture,
gravity and thermal gradients). Considering section X-X at 500 s, Simulation 2 yields a constant
capillary potential of !0·5 m which is evident to a distance of 32·5 mm. The capillary potential
from this point decreases (i.e. becomes less negative) to a value of !0·3 m toward X"50 mm at
the boundary signifying infiltration of moisture from this boundary. The section taken at 27·7 h
for simulation 2 shows an almost linear variation of capillary potential from the !0·3 m
boundary at X"50 mm to the !0·5 m boundary at X"0. The decrease in capillary potential
is evident as the infiltration of moisture from X50 has progressed across the domain with time.
Simulation 3, which includes the effects of a moisture gradient, gravity and a thermal gradient,
shows that the capillary potential has been decreased at a greater rate due to the inclusion of the
thermal gradient in the ½ dimension at both times. Again this has the effect of increasing the
moisture content at this position as moisture is driven away from the ½50 face.
Figure 6(c) shows capillary potential contours at plane cross section Z25 for Simulation 3. At
face ½50 large capillary potentials of up to ("!2·3 m are experienced due to the drying effect
of the elevated temperature. The curved nature of the contours reflects the imposed capillary
potential gradient in the X direction. The contour representing—0·5 m penetrates from ½50 to
½"36 mm indicating that moisture has been driven to the cooler areas of the domain.
5.4. Application 3—heat, moisture and air transfer application

This application illustrates the distributions, caused by a simulated heat source, of the three
primary variables of t, ¹ and P . This analysis is run as a fully coupled heat, moisture and air
!
transfer problem. Initial conditions are set as !0·5 m, 20°C and 100 kPa for capillary potential,
temperature and air pressure respectively. One element in the domain bordering external faces
X0, Z0 and ½0 (shaded in Figure 4(a)) has all nodes fixed at an elevated temperature of 60°C to
simulate a heater. All other external boundaries are non-conducting.
Results presented in Figures 7(a) and 7(b). Figure 7(a) shows the transient variation of capillary
potential at various points along section A-A (defined in Figure 4(c)). A-A varies in the X dimen-
sion at constant values of ½"0·0 mm and Z"12·5 mm. Temperature is also represented on
Figure 7(a) as a transient variation at point 25, 12·5, 0·0 (X, Z, ½) and is presented on the figure
using solid markers. The capillary potential variation at point (6·25, 12·5, 0·0) starts at the initial
value of !0·5 m and then decreases to below !0·46 m at 2500 s. This is caused by moisture
being driven from the heater initially increasing the moisture content at this position. From
2500 s the capillary potential then increases to just over !0.485 m at 10 000 s. Considering the
temperature at this time is approaching its maximum at this point the moisture content decreases
as water is pushed to the further extremes of the cube. This can be seen by the corresponding
minimum value of capillary potential of !0·422 m for point (42·75, 12·5, 0·0). From 10,000 s the
capillary potential for point (6·25, 12·5, 0·0) then decreases to an equilibrium position at approx-
imately !0·45 m after 75000 s. As the domain has nearly all reached a maximum temperature of
60°C moisture is no longer driven from the heater and therefore redistributes uniformly around
the cube. This is also evident from point (42·75, 12·5, 0·0). The reason that the capillary potential
has not returned to its initial value of !0·5 m is due to the fact that air pressure throughout the
domain changes uniformly with time.
A qualitative temperature distribution is shown in Figure 7(b) over the whole domain after 100 s.
Again red represents 60°C and blue represents 20°C. The external faces of each element are shown in
the figure and have been ‘shrunk’ to allow an internal view of the cube. The progress of the thermal
front into the cube is spherical in shape with the origin being at point 0.50, 0 as would be expected.
6. CONCLUSION
A new three-dimensional numerical model of coupled heat, moisture and air transfer in un-
saturated soil has been presented for the first time in the literature. The model accommodates the
following thermophysical processes:
(i) moisture transfer in the form of liquid and vapour flow,
(ii) heat transfer arising from conduction, convection and latent heat of vaporisation, and
(iii) the bulk flow of dry air and the movement of air in a dissolved state.
A numerical solution of the three-dimensional theoretical formulation was achieved using the
finite element method to achieve spatial discretisation and the finite difference method to
accommodate the time-varying nature of the problem. In particular, spatial discretization is
Figure 7. (a) Application 3—Transient variation of capillary potential and temperature at various points along the
section A-A (b) Application 3—Temperature variation over the domain after 100 s
achieved using standard 20-node isoparametric brick elements. Numerical integration is achieved
using a 3]3]3 Gaussian method. Time marching is achieved via the use of a fully implicit
mid-interval backward difference scheme.
The theoretical basis of the model has been validated against experimental results elsewhere.
Therefore attention has been focused in this paper on verification of the new software. This has
been achieved via comparisons with independent solutions of heat, moisture and air transfer in an
unsaturated soil. Excellent correlation of results has been achieved thus providing confidence in
the new model.
The new model has been applied to a number of test cases which illustrate the need for the
development of a model which can fully include three dimensional behaviour. In particular, three
applications have been presented each of increasing complexity. The first application illustrates
three-dimensional heat transfer. This particular application has been successfully verified against
existing commercial finite element software.
In conclusion, a new three-dimensional model for coupled heat, moisture and air transfer has
been developed. The model has been verified against independent solutions A number of
applications of the model have illustrated how the coupled processes under consideration may
combine to yield complex three-dimensional problems even within a simple geometric domain.
Visualization of three-dimensional results has also been addressed.
NOTATION
C [g#[H !1]h ]X
!! # - !!
C [g#[H !1]h]X #[H !1]o (Lh /L¹ )D(
!T # !T # $! -
C( [g#[H !1]h ]X #[H !1]o (Lh /L¹ )D(
! # - !T # $! -
C ( o (Lh /L()D
L - - T
C o (Lh /L¹ )D(
LT - -
C specific heat capacity of dry air
1$!
C specific heat capacity of soil water
1-
C specific heat capacity of soil solids
14
C specific heat capacity of soil vapour
17
C ( C (¹!¹ )C (!C (o /o )(¹!¹ )C (#C (¹!¹ )
T o- 3 L o$! $! - 3 L o$! 3
(g!h )X (#[C (¹!¹ )#¸]C (
- ! o7 3 7
C H#C (¹!¹ )C !C (o /o )(¹!¹ )#C (¹!¹ )(g!h )X
TT o- 3 LT o$! $! - 3 o$! 3 - !T
#[C (¹!¹ )#¸]C
o7 3 7T
C C (¹!¹ )(g!h )X
T! 1$! 3 - !!
C ( [g!h ](o gh/R ¹)!o H(Lh /L()D
V - 0 7 0 - T
C ![g!h ](o (gh/R ¹2)#[g!h ]hb!o h(Lh /L¹ )D(
VT - 0 7 - 0 -
C( C (#C (
t L V
C( C #C
T LT VT
D molecular diffusity of water vapour in air
!5.
E evaporation term (moisture transfer between liquid and vapour phases)
F approximate flux of heat normal to the boundary surface
)9(9/-,5)
g acceleration due to gravity
H volumetric heat capacity of unsaturated soil
H Henry’s constant
#
h relative humidity
K unsaturated hudraulic conductivity
K intrinsic permeability of the air phase
!
K [o ] (k K /k )#(o H K/c )
!! $! #0/45 ! 3! ! $! # 8
K( [o ] H K
! $! #0/45 #
K K/g
L!
K ( oK
L -
K [C o (¹!¹ )#C o (¹!¹ )#o ¸](k K /k )
T! 17 7 3 1$! $! 3 7 ! 3! !
#[C Ko (¹!¹ )/c ]
1- - 3 8
K j#C o (¹!¹ )K #o ¸K
TT 17 - 3 72 - 72
K ( C (¹!¹ )K (#C o (¹!¹ )K #o ¸K
T 1- 3 L 17 - 3 7- - 7-
K ![D lgo gh/o R ¹ ]
V- !5. 0 - 7
K !D lg(+¹ ) /o +¹ [hb!o (gh/R ¹ 2]
V2 !5. ! - 0 7
K (o ) (K k /k )
V! 7 #0/45 ! 3! !
K [D lg(+¹ ) /+¹ ][hb!(o (gh/R ¹ 2)]
VT !5. ! 0 7
K ( D lgo gh/R ¹
V !5. 0 7
K( K #K
! L! 7!
K( K
T VT
K(( K (#K (
L V
k relative permeability of air phase
3!
¸ latent heat of vapourisation of water
N,N matrix shape functions associated with element nodal points
3 4
n porosity
P air pressure
!
P pore liquid pressure
-
R specific gas constant for dry air
$!
R specific gas constant for water vapour
7
¹ absolute temperature
¹ reference temperature
3
(+¹ )
! ratio of microscopic temperature gradient in pore space to macroscopic temperature
+¹
t time
º liquid velocity
» vapour velocity
» air velocity
!
» ![» #H º]X
!! ! # !!
» ![» #H º]X
!T ! # !T
»( ![» #H º]X (
! ! # !
» Volume of soil liquid
-
» Volume of soil solids
4
» !C » (¹!¹ )X
T! 1$! ! 3 !!
G H
» C o º#C o »#C o » #C o » #C » (¹!¹ )X #
TT ! 1- - 17 - 17 7 ! 1$! $! ! 1$! ! 3 !T
C (¹!¹ )» [hb!(o (gh/R ¹ 2)]#¸» [hb!(o (gh/R ¹ 2)]
17 3 ! 0 7 ! 0 7
» ( ![C (¹!¹ )» (o gh/R ¹ )#¸» (o gh/R ¹ )#C (¹!¹ )» X (]
T 17 3 ! 0 7 ! 0 7 1$! 3 ! !
» !» [hb!(o (gh/R ¹ 2)]
VT ! 0 7
»( !» o gh/R ¹
V ! 0 7
»( »
T VT
»(( »(
V
X( o gh/R ¹
! 0 $!
X ![P /R ¹ 2#R /R [hb!o (gh/R ¹ 2]]
!T ! $! 7 $! 0 7
X 1/R ¹
!! $!
x, z, y global co-ordinate system

m, g, f local co-ordinate system
b Lo /L¹
0
c specific weight of water (c "o g)
- - -
h volumetric liquid content!h "» /»
- - - 4
j thermal conductivity
k dynamic viscosity of water
k viscosity of the air
!
l mass flow factor
o density of saturated water vapour
0
o density of liquid water
-
o density of dry air
$!
o density of soil particles
4
o density of water vapour
7
( capillary potential!("(P !P )/c
- ! -
REFERENCES
1. H. R. Thomas and M. R. Sansom, (1995) ‘Fully coupled analysis of heat, moisture, and air transfer in unsaturated soil.’
ASCE J. Engng. Mech., Vol. 121, March, 1995, 392—405.
2. S. D. King ‘A potential based model of coupled heat and moisture transfer in unsaturated soil.’ Ph.D ¹hesis.
University of Wales, Cardiff, (1991).
3. H. R. Thomas and S. D. King ‘Coupled temperature/capillary potential variations in unsaturated soil.’ J. Engng.
Mech., 117, 2475—2491 (1991).
4. J. R. Philip and D. A. DeVries, ‘Moisture movement in porous materials under temperature gradients.’ ¹rans. Am.
Geophysical ºnion. 38, 222—232 (1957).
5. H. R. Thomas and S. D. King, ‘Coupled heat and mass tranfer in unsaturated soil—a potential-based solution.’ Int. J.
Num. Anal. Meth. Geomech., 16, 757—773 (1992).
6. C. L. W. Li, ‘The use of transputers to simulate heat and mass transfer in soils.’ Ph.D. ¹hesis. University of Wales,
Cardiff, (1992).
7. M. R. Sansom, ‘A fully coupled numerical model of moisture, air and heat transfer in unsaturated soil.’ Ph.D. ¹hesis.
University of Wales, Cardiff, 1995.
8. O. C. Zienkiewicz and R. L. Taylor, (1989) ¹he Finite Element Method, McGraw Hill, 4th ed.
9. J. Ewen, ‘Combined heat and mass transfer in unsatuturated sand surrounding a heated cylinder.’ Ph.D. ¹hesis,
University of Wales, Cardiff, (1987).
10. N. E. Edlefsen and A. B. C. Anderson, ‘Thermodynamics of soil moisture.’ Hilgardia, 15(2), 31—298 (1943).
11. L. Barden, ‘Consolidation of compacted and unsaturated clays.’ Geotechnique, London, England 15(3), 267—286
(1965).
12. E. E. Alonso, F. Batle, A. Gens and A. Lloret, ‘Consolidation analysis of partially saturated soils—application to
earthdam construction.’ Num. Meth. In Geomech., Innsbruck, Austria, 1303—1308 (1988).
13. R. C. Weast, Handbook of chemistry and physics, 57th Ed., CRC Press, Cleveland, Ohio (1976).
14. D. W. Pollock, ‘Simulation of fluid flow and energy process associated with high-level radioactive waste disposal in
unsaturated alluvium.’ ¼ater Resour. Res., 22, 765—775 (1986).
15. O. C. Zienkiewicz and K. Morgan, Finite elements and approximation, John Wiley and Sons, New York (1983).
16. B. Irons and S. Ahmad, ¹echniques of finite elements, Ellis Horward Ltd, Chichester, U.K.(1980).
17. J. Ewen and H. R. Thomas, The thermal probe—a new method and its use on an unsaturated sand.’ Geotechnique,
¸ondon, England, 37, 91—105 (1987).
18. J. Ewen and H. R. Thomas ‘Heating unsaturated medium sand.’ Geotechnique, London, England, 39, 455—470 (1989).
19. PAFEC (1984) Program for Automatic Finite Element Calculations PAFEC Ltd. Nottingham.
20. A. D. Jefferson, ‘Finite element analysis of composite structures’ PhD ¹hesis. ºniversity of ¼ales, Cardiff, (1989).
21. H. R. Thomas ‘Non-linear analysis of heat and moisture transfer in unsaturated soils.’ J. Engng. Mech., 113,
1163—1180 (1987).

Thomas 1998

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Thomas 1998

Uploaded by

Copyright:

Available Formats

INTERNATIONAL JOURNAL FOR NUMERICAL AND ANALYTICAL METHODS IN GEOMECHANICS, VOL.

22, 75—95 (1998)

THREE-DIMENSIONAL HEAT, MOISTURE AND AIR

H. R. THOMAS*, S. W. REES AND N. J. SLOPER

CCC 0363—9061/98/020075—21$17.50 Received 25 July 1996

2.1. The governing differential equation for moisture flow

#» +(#» +¹#o E (2)

#» +(#» ¹#+[K +z] (3)

2.2. The governing differential equation for dry air flow

2.3. The governing differential equation for heat flow

3. THE NUMERICAL SOLUTION OF THE GOVERNING PARTIAL DIFFERENTIAL

Figure 1. Boundary flux on a 20 noded isoparametric ‘brick’ element

P) [KT +N3+N4 ] d)%W4#P) [KTT+N3+N4 ] d)%T4#P) [KT!+N3+N4 ] d)%P!4

P) [K! +N3+N4 ] d)%W4#P) [K!4+N3+N4 ] d)%P!4!P) [N3»! +N4 ] d)% W4

Figure 2. Verification—Column finite element meshes: (a) 2D; (b) 3D

5.1. Finite element discretisation

5.2. Application 1—Thermal application

5.3. Application 2—Heat and moisture transfer application

Table I. Soil properties for Leighton Buzzard Sand

Density (o ) 2700 kg/m3

Dirichlet boundary conditions

Simulation Gravity Capillary potential (() Temperature (¹ )

5.4. Application 3—heat, moisture and air transfer application

x, z, y global co-ordinate system

You might also like